Trade Theory with Numbers: Quantifying the Welfare Consequences of Globalization Andrés Rodríguez-Clare (UC Berkeley and NBER) September 29, 2012
The Armington Model
The Armington Model CES preferences: This implies that U j = ( ) σ/(σ 1) Q (σ 1)/σ ij, σ > 1 i X ij = P1 σ ij k P 1 σ kj }{{} λ ij Y j }{{} total income
The Armington Model If wages are w i and (iceberg) trade costs are τ ij, then P ij = w i τ ij, so X ij = (w i τ ij ) 1 σ k (w k τ kj ) 1 σ Y j This is the Gravity Equation for the Armington model. We refer to σ 1 as the trade elasticity. Technically, d ln X ij /X jj d ln τ ij = σ 1
Armington: Trade and Welfare The expenditure function with CES preferences is e(p 1, P 2,..., P N, U) = ( i P 1 σ i ) 1/(1 σ) } {{ } P = "price index" U Expenditure per capita in country j is Y j /L j = w j, hence w j = P j U j and so U j = w j /P j. We can think of this as "real income in country j" We want to know how changes in trade costs affect w j /P j
Armington: Trade and Welfare Come back to Let Since P ij = w i τ ij and P 1 σ j λ ij = X ij = λ ij P1 σ ij k P 1 σ kj Y j X ij k X kj = X ij Y j = k P 1 σ kj P1 σ ij k P 1 σ kj then = (w i τ ij ) 1 σ P 1 σ j
Armington: Trade and Welfare So we have Since τ jj = 1, then λ ij = (w i τ ij ) 1 σ P 1 σ j λ jj = w j 1 σ Pj 1 σ = ( ) 1 σ wj P j Finally, U j = w j = λ 1/(1 σ) jj P j
Armington: Gains from Trade How bad would it be to shut down trade? In autarky have λ jj = 1, hence U A j /U j = λ 1/(σ 1) jj Define gains from trade as GT j 1 U A j /U j = 1 λ 1/(σ 1) jj Note that ε σ 1 is the trade elasticity, so can write GT j = 1 λ 1/ε jj
Armington: Gains from Trade To implement GT j = 1 λ 1/ε jj, need to measure λ jj and need an estimate of ε
Armington: Gains from Trade To implement GT j = 1 λ 1/ε jj, need to measure λ jj and need an estimate of ε λ jj is share of gross expenditure devoted to home purchases, computed from OECD-STAN as gross production - exports gross expenditure
Armington: Gains from Trade To implement GT j = 1 λ 1/ε jj, need to measure λ jj and need an estimate of ε λ jj is share of gross expenditure devoted to home purchases, computed from OECD-STAN as gross production - exports gross expenditure To estimate ε, return to gravity equation, X ij = (w i τ ij ) ε Pj ε Y j Taking logs, we have ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij If we had data for τ ij, then can run OLS regression of ln X ij on ln τ ij with exporter and importer dummies to estimate ε
Armington: Gains from Trade To implement GT j = 1 λ 1/ε jj, need to measure λ jj and need an estimate of ε λ jj is share of gross expenditure devoted to home purchases, computed from OECD-STAN as gross production - exports gross expenditure To estimate ε, return to gravity equation, X ij = (w i τ ij ) ε Pj ε Y j Taking logs, we have ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij If we had data for τ ij, then can run OLS regression of ln X ij on ln τ ij with exporter and importer dummies to estimate ε A central estimate of ε is 5
Armington: Gains from Trade We have ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij + ξ ij
Armington: Gains from Trade We have How do we measure τ ij? ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij + ξ ij
Armington: Gains from Trade We have How do we measure τ ij? Tariffs ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij + ξ ij
Armington: Gains from Trade We have How do we measure τ ij? Tariffs Price differences ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij + ξ ij
Armington: Gains from Trade We have How do we measure τ ij? Tariffs Price differences A central estimate of ε is 5 ln X ij = ε ln w i + ln ( Y j Pj ε ) ε ln τij + ξ ij
Armington: Gains from Trade λ jj % GT j Canada 0.87 2.7 Denmark 0.89 2.2 France 0.91 1.8 Portugal 0.84 3.3 Slovakia 0.79 4.6 U.S. 0.95 1.0
Armington: Counterfactuals More generally, us implies that (notation ˆx = x /x) U j = w j = λ 1/ε jj P j Û j = λ 1/ε jj
Armington: Counterfactuals More generally, us implies that (notation ˆx = x /x) U j = w j = λ 1/ε jj P j Û j = λ 1/ε jj For ex-ante analysis, need to predict ˆλ jj
Armington: Counterfactuals More generally, us implies that (notation ˆx = x /x) U j = w j = λ 1/ε jj P j Û j = λ 1/ε jj For ex-ante analysis, need to predict ˆλ jj Two approaches
Armington: Counterfactuals More generally, us implies that (notation ˆx = x /x) U j = w j = λ 1/ε jj P j Û j = λ 1/ε jj For ex-ante analysis, need to predict ˆλ jj Two approaches Diffi cult: calibrate model and simulate counterfactual
Armington: Counterfactuals More generally, us implies that (notation ˆx = x /x) U j = w j = λ 1/ε jj P j Û j = λ 1/ε jj For ex-ante analysis, need to predict ˆλ jj Two approaches Diffi cult: calibrate model and simulate counterfactual Easy: use data { λ ij, Y j }, ε and { ˆτij } to get implied ˆλ jj
Cheese, really?
Cheese, really? What about comparative advantage? (Ricardo, Eaton-Kortum)
Cheese, really? What about comparative advantage? (Ricardo, Eaton-Kortum) Variety gains? (Krugman)
Cheese, really? What about comparative advantage? (Ricardo, Eaton-Kortum) Variety gains? (Krugman) Selection/reallocation effects? (Melitz)
Cheese, really? What about comparative advantage? (Ricardo, Eaton-Kortum) Variety gains? (Krugman) Selection/reallocation effects? (Melitz) Pro-competitive effects? (Melitz-Ottaviano)
Comparative Advantage (Ricardo, Eaton-Kortum) Another reason to trade beyond cheese: specialization according to comparative advantage
Comparative Advantage (Ricardo, Eaton-Kortum) Another reason to trade beyond cheese: specialization according to comparative advantage In Ricardian theory, comparative advantage arises because of differences in technology
Comparative Advantage (Ricardo, Eaton-Kortum) Another reason to trade beyond cheese: specialization according to comparative advantage In Ricardian theory, comparative advantage arises because of differences in technology Cheese + Ricardo?
Comparative Advantage (Ricardo, Eaton-Kortum) Trade elasticity now comes from productivity rather than preferences (ε = θ rather than ε = σ 1)
Comparative Advantage (Ricardo, Eaton-Kortum) Trade elasticity now comes from productivity rather than preferences (ε = θ rather than ε = σ 1) But welfare implications of an international shock are still Ŵ j = λ 1/ε jj
Comparative Advantage (Ricardo, Eaton-Kortum) Trade elasticity now comes from productivity rather than preferences (ε = θ rather than ε = σ 1) But welfare implications of an international shock are still Ŵ j = λ 1/ε jj Given same data {λ ij, Y j }, same ε, and same shock { ˆτ ij }, then same ˆλ jj and same Ŵ j
Increasing Returns with Differentiated Goods (Krugman) In principle, gains from increasing returns and more variety
Increasing Returns with Differentiated Goods (Krugman) In principle, gains from increasing returns and more variety In practice, with CES preferences, both variety and scale unchanged
Increasing Returns with Differentiated Goods (Krugman) In principle, gains from increasing returns and more variety In practice, with CES preferences, both variety and scale unchanged Gains arise from cheaper foreign goods, just as in Armington
Increasing Returns with Differentiated Goods (Krugman) In principle, gains from increasing returns and more variety In practice, with CES preferences, both variety and scale unchanged Gains arise from cheaper foreign goods, just as in Armington Gravity and gains as in the Armington model
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms Trade > higher average effi ciency
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms Trade > higher average effi ciency But what about welfare?
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms Trade > higher average effi ciency But what about welfare? First problem: in Melitz, heterogeneity in productivity, but same VMPL across firms
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms Trade > higher average effi ciency But what about welfare? First problem: in Melitz, heterogeneity in productivity, but same VMPL across firms Second problem: exit of low productivity firms entails a loss of variety
Reallocation Effects (Melitz) Fixed exporting costs > only highly productive firms export Trade > reallocation of labor from low to high productivity firms Trade > higher average effi ciency But what about welfare? First problem: in Melitz, heterogeneity in productivity, but same VMPL across firms Second problem: exit of low productivity firms entails a loss of variety Bottom line: Melitz model (with Pareto) leads to same formula, Ŵ j = λ 1/ε jj
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups But this happens only when economy is close to autarky
Pro-competitive Effects? Key idea: stronger competition weakens domestic monopolies, improving effi ciency But: lower tariffs allow foreign monopolists to increase their markups Which effect dominates? It depends on the model Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups But this happens only when economy is close to autarky For open economies, effects are small
Pro-competitive Effects? Two ways to go: drop MC or drop CES
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups But this happens only when economy is close to autarky
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Calibrate to Taiwanese firm-level data There can be large pro-competitive effects of trade through destruction of monopolies Formally, trade reduces dispersion in markups But this happens only when economy is close to autarky For open economies, effects are small
Pro-competitive Effects? Two ways to go: drop MC or drop CES
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Arkolakis, Costinot, Donaldson and RC drop CES but keep Pareto (to get gravity)
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Arkolakis, Costinot, Donaldson and RC drop CES but keep Pareto (to get gravity) Markup distribution not affected: incumbents lower markups, but low markup firms leave
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Arkolakis, Costinot, Donaldson and RC drop CES but keep Pareto (to get gravity) Markup distribution not affected: incumbents lower markups, but low markup firms leave Non-homothetic preferences imply that a decline in trade costs reallocates demand towards low markup goods
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Arkolakis, Costinot, Donaldson and RC drop CES but keep Pareto (to get gravity) Markup distribution not affected: incumbents lower markups, but low markup firms leave Non-homothetic preferences imply that a decline in trade costs reallocates demand towards low markup goods Pro-competitive effects actually lower gains from trade
Pro-competitive Effects? Two ways to go: drop MC or drop CES Edmon-Midrigan-Xu drop MC and use Cournot competition Arkolakis, Costinot, Donaldson and RC drop CES but keep Pareto (to get gravity) Markup distribution not affected: incumbents lower markups, but low markup firms leave Non-homothetic preferences imply that a decline in trade costs reallocates demand towards low markup goods Pro-competitive effects actually lower gains from trade Effect is very small
Back to Armington 1. Add multiple sectors 2. Add traded intermediates
Multiple sectors, GT Upper level EoS ρ and lower level EoS ε s Recall gains for Canada of 2.7%. Now gains can be much higher: ρ = 1 implies GT = 8%
Multiple sectors, GT, role of upper level EoS
Tradable intermediates, GT Set ρ = 1, add tradable intermediates with Input-Output structure Labor shares are 1 α j,s and input shares are α j,ks ( k α j,ks = α j,s )
Tradable intermediates, GT % GT j % GT MS j % GT IO j Canada 2.7 8.0 19.2 Denmark 2.2 6.1 15.3 France 1.8 3.5 10.5 Portugal 3.3 9.1 26 U.S. 1.0 1.8 4.4
Combination of micro and macro features In Krugman, free entry leads to scale effects associated with total sales
Combination of micro and macro features In Krugman, free entry leads to scale effects associated with total sales In Melitz, additional scale effects associated with market size
Combination of micro and macro features In Krugman, free entry leads to scale effects associated with total sales In Melitz, additional scale effects associated with market size In both models, trade may affect entry and fixed costs
Combination of micro and macro features In Krugman, free entry leads to scale effects associated with total sales In Melitz, additional scale effects associated with market size In both models, trade may affect entry and fixed costs All these effects do not play a role in the one sector model
Combination of micro and macro features In Krugman, free entry leads to scale effects associated with total sales In Melitz, additional scale effects associated with market size In both models, trade may affect entry and fixed costs All these effects do not play a role in the one sector model With multiple sectors and traded intermediates, these effects come back
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 Key: Cobb-Douglas preferences
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 Key: scale effects linked to entry and L.O.V., compositional effect
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 Key: IO loops and interdependencies
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7 Key: further amplification through trade in capital goods
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7 MS, IO, MC, η s = 0 25.1 23.2 16.8 10.3 5.8
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7 MS, IO, MC, η s = 0 25.1 23.2 16.8 10.3 5.8 Key: trade expands markets relative to entry cost > entry
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7 MS, IO, MC, η s = 0 25.1 23.2 16.8 10.3 5.8 MS, IO, MC, η s = 1.5 33.6 44.8 25.5 14.9 10.1
Gains from Trade... Canada China Germany New Zealand US Aggregate 2.7 0.8 2 2.5 1 MS, PC 8 1.1 3.7 6.9 1.8 MS, MC 8.9-0.2 4.6 2.1 1.3 MS, IO, PC 19.2 5.9 9.8 14.3 4.4 MS, IO(k), PC 26.8 9.3 14.3 19.8 6.7 MS, IO, MC, η s = 0 25.1 23.2 16.8 10.3 5.8 MS, IO, MC, η s = 1.5 33.6 44.8 25.5 14.9 10.1
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj This is what CGE exercises do
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj This is what CGE exercises do Contribution here:
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj This is what CGE exercises do Contribution here: Link to theory "mid-sized models"
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj This is what CGE exercises do Contribution here: Link to theory "mid-sized models" Model consistent estimation
From GT to trade policy evaluation Back to {λ ij, Y j }, ε and { ˆτ ij } to get implied ˆλ jj This is what CGE exercises do Contribution here: Link to theory "mid-sized models" Model consistent estimation Quantify mechanisms
Fit of model Research agenda
Research agenda Fit of model Elasticities
Research agenda Fit of model Elasticities Scale Effects
Final Thoughts Complementarities between trade and MP/diffusion can increase GT
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich?
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich? Gains from openness + domestic frictions
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich? Gains from openness + domestic frictions Scale economies = y DNK /y US = 1/3
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich? Gains from openness + domestic frictions Scale economies = y DNK /y US = 1/3 Domestic frictions bring this to 2/3
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich? Gains from openness + domestic frictions Scale economies = y DNK /y US = 1/3 Domestic frictions bring this to 2/3 Relative GO are then 3/2
Final Thoughts Complementarities between trade and MP/diffusion can increase GT Gains from openness (including MP, diffusion) can be much larger How can small countries be rich? Gains from openness + domestic frictions Scale economies = y DNK /y US = 1/3 Domestic frictions bring this to 2/3 Relative GO are then 3/2 If GO US = 5% (twice GT) then GO DNK 50%